Myers-Briggs

So this is really a compression algorithm. We take a medium sized input set, compress it to 4 bits and then we can use those 4 bits to answer lots of interesting question. And by combining personality types with context (work, parenting, relationships) we can make further predictions.

But I'm not sure the base model is solid. For instance (intro/extra/ambi)-version follows a spectrum that is a bell curve over the population. Binary classifications are more meaningful when they at least flatter or even divergent.

So what we really want is N dimensional space with regions defined so that people within a region share important personality characteristics and the boundaries between regions are distinct.